Looking for needles in haystacks
Here's a proper programming challenge, simple to state, and in the end quite challenging to solve. A function produces a 64-bit output given a 64-bit input. It mostly behaves like a random function, so given a random input, most outputs occur with probability between 0 and 2-63; in other words, for a randomly-chosen 64-bit y, there are usually between 0 and 2 solutions to the equation f(x) = y. However, a few occur with much greater frequency - more like 2-32 or so. The function is pretty quick and easy to calculate, but its internal structure defies direct analysis. How would you identify these more frequent occurrences by experiment? How big a computer would you need to do it in something like a few hours to a few days? Think about it for a while before you read on.